Federal Communications Commission Washington, D.C

Transcription

1 Federal Communications Commission Washington, D.C June 25, 2013 Marlene H. Dortch Secretary Federal Communications Commission th Street, SW Washington, DC Dear Ms. Dortch: Re: Connect America Fund, WC Docket No On Tuesday, October 9, 2012, the Wireline Competition Bureau (Bureau) announced the commencement of the Connect America Cost Model (CAM) virtual workshop to solicit input and facilitate discussion on topics related to the development and adoption of the forwardlooking cost model for Connect America Phase II. 1 To date, the Bureau has solicited comment on 28 different topics in the CAM virtual workshop. 2 On three occasions, the Bureau has submitted into the record comments posted by parties in the CAM virtual workshop. 3 These submissions reflect comments posted to the CAM virtual workshop through April 29, With this letter, the Bureau supplements the previous record by submitting comments posted by parties in the CAM virtual workshop from April 30, 2013 through June 24, In addition, various parties have discussed the potential relevance of International Telecommunication Union (ITU) standards and the Commission s Measuring Broadband America (MBA) program in the record. The Bureau may rely on ITU latency calculations, MBA data and other technical documents in implementing the Commission s latency requirement. We accordingly place the following sources into the record: 1 Wireline Competition Bureau Announces Commencement of Connect America Phase II Cost Model Virtual Workshop, WC Docket Nos , , Public Notice, 27 FCC Rcd (Wireline Comp. Bur. 2012). See also Wireline Competition Bureau Announces Connect America Phase II Cost Model Virtual Workshop, WC Docket Nos , , Public Notice, 27 FCC Rcd (Wireline Comp. Bur. 2012). 2 WCB Cost Model Virtual Workshop, (follow link to the Cost Model Virtual Workshop). 3 Letter from Michael Jacobs, Legal Advisor to the Chief, Wireline Competition Bureau, FCC, to Marlene H. Dortch, Secretary, FCC, WC Docket No (filed Feb. 6, 2013); Letter from Jamie Susskind, Legal Advisor to the Chief, Wireline Competition Bureau, FCC, to Marlene H. Dortch, Secretary, FCC, WC Docket No (filed Mar. 27, 2013); Letter from Michael Jacobs, Legal Advisor to the Chief, Wireline Competition Bureau, FCC, to Marlene H. Dortch, Secretary, FCC, WC Docket No (filed Apr. 30, 2013).

2 FED. COMMC NS COMM N OFFICE OF ENG G AND TECH. AND CONSUMER AND GOVERNMENTAL AFFAIRS BUREAU, 2013 MEASURING BROADBAND AMERICA FEBRUARY REPORT, A REPORT ON CONSUMER WIRELINE BROADBAND PERFORMANCE IN THE U.S. (2013), Broadband-America-feb-2013.pdf. FED. COMMC NS COMM N OFFICE OF ENG G AND TECH. AND CONSUMER AND GOVERNMENTAL AFFAIRS BUREAU, MEASURING BROADBAND AMERICA 2013 REPORT, VALIDATED DATA 2013 (2013), INTERNET ENG G TASK FORCE, FRAMEWORK FOR TCP THROUGHPUT TESTING (REQUEST FOR COMMENT 6349) (2011) INTERNET ENG G TASK FORCE NETWORK WORKING GROUP, A SIMPLE NETWORK MANAGEMENT PROTOCOL (SNMP) (1990), Internet Eng g Task Force Proposed Working Group, Large Scale Measurement of Access Network Performance (lmap), (last visited June 21, 2013). INT L TELECOMM. UNION, TELECOMM. STANDARDIZATION SECTOR, SERIES G: TRANSMISSION SYSTEMS AND MEDIA, DIGITAL SYSTEMS AND NETWORKS, G.114 (2003), INT L TELECOMM. UNION, TELECOMM. STANDARDIZATION SECTOR, SERIES Y: GLOBAL INFORMATION INFRASTRUCTURE, INTERNET PROTOCOL ASPECTS AND NEXT-GENERATION NETWORKS, Y.1541 (2011), Y.1541/en. Respectfully Submitted, /s/ Michael J. Jacobs Legal Advisor to the Chief Wireline Competition Bureau Attachment 2

3 Connect America Cost Model Virtual Workshop Questions and Comments Posted Between April 30, 2013 and June 25, 2013 Between April 30, 2013, and June 25, 2013, the Wireline Competition Bureau (Bureau) posted follow-up questions in the comment section for the following topics: DETERMINING THE FRACTION OF SUPPORTED LOCATIONS THAT WILL RECEIVE SPEEDS OF 6 MBPS/1.5 MBPS OR GREATER Background The USF/ICC Transformation Order requires price cap carriers that accept the state-level commitment for universal service support under Connect America Phase II to offer broadband at speeds of 4 Mbps/1 Mbps to all supported locations and at least speeds of 6 Mbps/1.5 Mbps to a number of supported locations by the end of the fifth year. The Commission directed the Wireline Competition Bureau to design the forward-looking cost model so that it ensures that the most locations possible receive broadband at speed of 6 Mbps/1.5 Mbps or greater at the end of the five year term. Question(s) 1) The ABC Coalition has argued that carriers that receive Connect America Phase II support will generally choose to build or maintain fiber-to-the-dslam networks rather than build new fiber-tothe-premises networks. How specifically should the Bureau determine what fraction of locations would reasonably be required to receive speeds of at least 6 Mbps/1.5 Mbps? Would it be appropriate to calculate the number of locations likely to receive speeds of 6 Mbps/1.5 Mbps when the network is engineered to deliver at least 4 Mbps/1 Mbps to the most distant supported locations? What assumptions should be made regarding the gauge of the copper and the maximum copper loop length? Comments Steve Rosenberg Due to the variation in cost among geographies, a uniform nationwide percentage for the number of locations to be served with 6 Mbps/1.5 Mbps service may not be a reasonable approach, as it could result in obligations that are too difficult to achieve in some areas, while failing to establish readily achievable deployment obligations in other areas for recipients of Phase II funding. Is varying the 6 Mbps/1.5 Mbps service obligation by geographic area the best approach? The ABC Coalition argues that price cap carriers receiving Phase II support will generally build fiber-tothe-digital subscriber line access multiplexer (DSLAM) networks, and that 4,000 feet is the maximum distance from a DSLAM over which a copper loop can be used to provide 6 Mbps/1.5 Mbps broadband service (see below Comments of Robert Mayer, USTelecom on Behalf of the ABC Coalition). The Connect America Cost Model (CAM) is capable of modeling both 5,000-foot and 12,000-foot DSL architectures. In order to establish the specific percentage of eligible locations to be served with 6 Mbps/1.5 Mbps service, should we use the CAM to estimate the number of eligible locations that are within 4,000 feet of a DSLAM in a 12,000-foot DSL architecture? Alternatively, should we use the CAM to estimate the number of eligible locations that are within 4,000 feet of a DSLAM in a 5,000-foot DSL architecture? Is this an appropriate approach for determining the percentage of eligible locations to be served with 6 Mbps/1.5 Mbps? Charts showing the percentage of eligible locations that would be required to be served with 6 Mbps/1.5 Mbps service under both architectures, based on recent model runs, can be viewed below. (A chart 1

4 showing the percentage of eligible locations for a 5,000-foot DSL architecture is attached to this comment and a chart showing the percentage of eligible locations for a 12,000-foot DSL architecture is attached to the comment below). If the Bureau were to use this methodology with an assumed 12,000-foot architecture, most carriers, in most states, would have an obligation to provide 6 Mbps/ 1.5 Mbps service to between 10 and 25 percent of eligible locations, although there would be some outliers outside that range. If the Bureau were to use this methodology with an assumed 5,000-foot architecture, most carriers would have an obligation to provide 6 Mbps/1.5 Mbps service to between 50 and 65 percent of eligible locations, although there would be some outliers outside that range as well. Which architecture should be assumed if we were to adopt such a methodology? If we were to use such a methodology, should we establish a minimum percentage of locations to be served with 6 Mbps/1.5 Mbps service that varies on a state-by-state basis? Should there be any flexibility under such an approach for a carrier accepting funding to satisfy its obligations by averaging at the holding company level the percentage of locations to be served with 6 Mbps/1.5 Mbps service for those states where it accepts Phase II support? When the Bureau adopts the final cost model, should it specify the number of locations, either by state or in the aggregate, that each carrier accepting support would be required to serve with 6 Mbps/1.5 Mbps service? The Bureau also invites commenters to propose other methodologies for determining the number of locations to be served with 6 Mbps/1.5 Mbps service. 2

5 This chart shows the percentage of eligible locations that would be required to be served with 6 Mbps/1.5 Mbps service under a 12,000-foot DSL architecture, based on recent model runs, as referenced in the comment above from Steve Rosenberg, Chief Data Officer, WCB. Robert Mayer, United States Telecom Association on behalf of ABC Coalition Due to the variation in cost among geographies, a uniform nationwide percentage for the number of locations to be served with 6 Mbps/1.5 Mbps service may not be a reasonable approach, as it could result in obligations that are too difficult to achieve in some areas, while failing to establish readily achievable deployment obligations in other areas for recipients of Phase II funding. QUESTION 1: Is varying the 6 Mbps/1.5 Mbps service obligation by geographic area the best approach? Response: The Coalition agrees that there are significant variations by state in the percentage of locations that can be served by a 6 Mbps/1.5 Mbps service when offered as an extension of a 4 Mbps/1.5 3

6 Mbps service. The Bureau s charts displaying the percentage of locations that are within 4,000 road distance feet of a DSLAM using 5kft and 12kft DSL Architecture confirms this variation. There are multiple means of recognizing this variation while implementing the 6 Mbps/1.5 Mbps standard. The Bureau could implement a state and company specific percentage, a state specific percentage, or develop a holding company percentage. The Coalition recommends a holding company specific percentage due to limitations in the rural high cost geographic data. The high cost areas where CAF 2 Model support is targeted contain very low density areas where geocoded locations are imprecise. For example, in the rural areas of the western U.S. road sections are a mile apart and many times only a handful of locations existing along that section. The site that the software places those locations has a very large implication as to whether the locations fall within the 4,000 road-feet of DSLAM or do not. This can have an impact on the calculated percentage by state by company. Having the flexibility to account for differences between the model and reality is necessary to ensure the goals of the CAF 2 program increasing the availability of broadband in high cost price cap areas. QUESTION 2: The Connect America Cost Model (CAM) is capable of modeling both 5,000-foot and 12,000-foot DSL architectures. In order to establish the specific percentage of eligible locations to be served with 6 Mbps/1.5 Mbps service, should we use the CAM to estimate the number of eligible locations that are within 4,000 feet of a DSLAM in a 12,000-foot DSL architecture? Response: The Coalition continues to state that the 12,000-foot DSL architecture is the correct configuration to utilize to estimate the number of eligible locations that can be served with 6 Mbps/1.5 Mbps service, given the 4 Mbps/ 1 Mbps standard that applies to all locations. QUESTION 3: Alternatively, should we use the CAM to estimate the number of eligible locations that are within 4,000 feet of a DSLAM in a 5,000-foot DSL architecture? Response: The Coalition disagrees that a 5,000-foot DSL architecture is appropriate for estimating the number of eligible locations that can be served with 6 Mbps/1.5 Mbps service for policy reasons. The Commission adopted an initial broadband speed benchmark for CAF recipients of 4 Mbps downstream and 1 Mbps upstream recognizing that some customers located closer to their serving device would be capable of receiving service in excess of the initial standard (par. 94 & fn 143, FCC ). This implies that the higher speed availability is incidental to the base speed of 4 Mbps downstream and 1 Mbps upstream to all locations. However, the Bureau s chart for the 5,000-foot DSL Architecture shows much higher percentages than can be considered incidental for eligible locations that are within 4,000 feet of the DSLAM. With an assumed 5,000-foot architecture, most price cap carriers would have an obligation to deploy the 6 Mbps/1.5 Mbps service to between 50 and 65 percent of eligible locations. This is far above the availability that the Commission s language and rules imply. If 50 to 65 percent of locations were required by the Bureau to be built to the 6 Mbps/1.5 Mbps standard, the Commission stated policy of the 4 Mbps/ 1 Mbps basic standard would effectively be changed. The Coalition believes that the 12,000-foot DSL architecture is more consistent with the intent of the Commission Order. QUESTION 4: Is this an appropriate approach for determining the percentage of eligible locations to be served with 6 Mbps/1.5 Mbps? Response: The Coalition continues to believe that using the CAM with a 12,000-foot DSL Architecture and determining the percentage of eligible locations within 4,000 road feet of the DSLAM is the best method to estimate the number of eligible locations to receive the 6 Mbps/1.5 Mbps service level. This methodology is grounded in the affirmation that the Coalition carriers will generally deploy fiber-to-the- DSLAM networks rather than build new fiber-to-the-node or fiber-to-the-premises networks. However, if the number/percentage of eligible locations to be served by 6 Mbps/1.5 Mbps service is determined by the 5,000-foot DSL architecture with its commensurate commitment to build 50 to 65 percent of eligible locations, many price cap carriers may turn down the right of first refusal for CAF 2. This result would not enhance the deployment of broadband to price cap rural areas and would in-fact hamper the expansion of broadband to unserved and underserved areas. The Bureau has provided charts showing the percentage of eligible locations that would be required to be served with 6 Mbps/1.5 Mbps service under both architectures, based on recent model runs. If the Bureau were to use this methodology with an assumed 12,000-foot architecture, most carriers, in most states, would have an obligation to provide 6 Mbps/ 1.5 Mbps service to between 10 and 25 percent of eligible 4

7 locations, although there would be some outliers outside that range. If the Bureau were to use this methodology with an assumed 5,000-foot architecture, most carriers would have an obligation to provide 6 Mbps/1.5 Mbps service to between 50 and 65 percent of eligible locations, although there would be some outliers outside that range as well. QUESTION 5: Which architecture should be assumed if we were to adopt such a methodology? Response: As stated above, the Coalition continues to believe that using the CAM with a 12,000-foot DSL Architecture and determining the percentage of eligible locations within 4,000 road feet of the DSLAM is the best method to estimate the number of eligible locations to receive the 6 Mbps/1.5 Mbps service level. QUESTION 6: If we were to use such a methodology, should we establish a minimum percentage of locations to be served with 6 Mbps/1.5 Mbps service that varies on a state-by-state basis? Response: Please see the response for question 1. QUESTION 7: Should there be any flexibility under such an approach for a carrier accepting funding to satisfy its obligations by averaging at the holding company level the percentage of locations to be served with 6 Mbps/1.5 Mbps service for those states where it accepts Phase II support? Response: Please see the response for question 1. QUESTION 8: When the Bureau adopts the final cost model, should it specify the number of locations, either by state or in the aggregate, that each carrier accepting support would be required to serve with 6 Mbps/1.5 Mbps service? Response: Please see the response for question 1. QUESTION 9: The Bureau also invites commenters to propose other methodologies for determining the number of locations to be served with 6 Mbps/1.5 Mbps service. Response: Please see the response for question 5. 5

8 RATE OF RETURN FOR CONNECT AMERICA COST MODEL Background The Commission previously adopted a unitary rate of return for all incumbent LECs regardless of size when such carriers were operating as regulated monopolies. Since that time, Congress enacted the 1996 Act, technology changes have introduced alternatives to the incumbent s service, and most of the larger incumbent LECs have moved to price cap regulation. Hybrid Cost Proxy Model: The authorized federal rate of return has been percent since January 1, HCPM utilizes the authorized rate of return. When establishing criteria to ensure consistency in the calculations of federal universal service support, the Commission concluded that the authorized federal rate of return on interstate services would be a reasonable rate of return by which to determine forward looking costs. Connect America Cost Model: The Connect America Cost Model (CACM) utilizes Annual Charge Factors (ACFs) to capture the annual cost of capital investments that are used over time, including the cost of initial deployment, replacement capital expense, and the cost of money necessary to have access to that amount of capital. The ABC Coalition previously submitted into the record a model that assumed a nine percent rate of return when calculating ACFs. In response to the June 2012 Model Design Public Notice, the American Cable Association submitted data about a number of price cap carriers, arguing that an appropriate rate of return for the price cap companies receiving model-based support should be lower than nine percent. The current version of the model assumes a nine percent rate of return in setting the ACFs. The model also assumes a debt-to-equity ratio of 25:75 to calculate taxes only on equity. Question(s) 1) In order to adopt final values for ACFs, the Bureau will need to make an assumption about the cost of money for price cap carriers receiving model-based support. What rate of return should the Bureau use in setting the ACFs when it adopts the final version of the model? Would it be appropriate for the final version of CACM to have different cost of money assumptions for different price cap carriers? Or, would the administrative benefits of using a single rate of return for all price cap carriers outweigh the complexity of establishing carrier-specific assumptions regarding appropriate rates of return? 2) Given the different regulatory treatment of price cap carriers, particularly for purposes of calculating universal service support, should the Commission use a different rate of return for purposes of calculating universal service support in the CACM for price cap carriers than the generally-authorized rate of return for incumbent rate-of-return carriers? Comments Steve Rosenberg Prior versions of the Connect America Cost Model (CAM) assumed a nine percent cost of capital in setting Annual Charge Factors (ACFs). The Wireline Competition Bureau (Bureau) released a staff report yesterday that determined that a reasonable analytical approach would establish a zone of reasonableness for the cost of capital of approximately seven percent to approximately nine percent calculated with a debt to equity ratio based on the market value of carriers capital structure. A Public Notice seeking comment on this staff report was also released yesterday (See Prescribing the Authorized Rate of Return: Analysis of Methods for Establishing Just and Reasonable Rates for Local Exchange Carriers, WC Docket No , Staff Report, DA (Wireline Comp. Bur. rel. May 16, 2013); Wireline Competition Bureau Seeks Comment on Rate of Return Represcription Staff Report, WC Docket No et al., Public Notice, DA (Wireline Comp. Bur. rel. May 16, 2013)). 6

9 CAM v provides users the option of selecting ACFs that assume a nine percent cost of capital, calculated with a ratio of debt to equity of 25:75, or an eight percent cost of capital, calculated with a ratio of debt to equity of 45:55. CAM v has two input collections. These input collections contain identical default input values, except one includes values for ACFs calculated with a nine percent cost of capital and the other includes values for ACFs calculated with an eight percent cost of capital. Parties who have signed the Third Supplemental Protective Order will be able to access these inputs collections by accessing the model, visiting the Posted Data Sets page, and viewing the ICCQA CAM312ACF8SBI6VoiceCblVoiceFW2 and ICCQA CAM312ACF9SBI6VoiceCblVoiceFW2 ZIP files under Model Inputs. Parties will also be able to view the impact on cost estimates of using different assumptions about the cost of capital and debt structure by accessing the model ( visiting the Posted Data Sets page and opening the Model Outputs labeled SS2013MMDDCAM312ACF9VoiceCablePlusVoiceFW2 and SS2013MMDDCAM312ACF8VoiceCablePlusVoiceFW2. The Third Supplemental Protective Order is available athttp:// For purposes of finalizing input values in the Connect America Cost Model, should the Bureau utilize an assumed cost of capital of eight percent, calculated with a ratio of debt to equity of 45:55, and a cost of debt of 6.19 percent, when adopting final annual charge factors? Would this be a reasonable reflection of the cost of capital and capital structure of an efficient provider? We note that carriers that have higher debt tend to have a higher cost of capital so that the added benefit of more tax shielding is generally at least somewhat offset by greater costs for both debt and equity. Robert Mayer, United States Telecom Association on behalf of ABC Coalition The Coalition responds to the new question posed in this Virtual Workshop topic on May 17, 2013, select portions of the May 16, 2013 Staff Report[1] and certain statements previously made by the American Cable Association ( ACA ) in its May 2, 2013 Ex Parte communication on this topic. In the comments below, we conclude that the Staff Report, when restricted to price cap carriers[2] and weighted by total capitalization for its analysis would yield a zone of reasonableness for cost of capital of 7.43% to 9.52%.[3] The Staff Report recommended that the Commission consider establishing the authorized rate of return in the upper half of the range, noting, among other things, the current historically low interest rates and the Commission s infrequency of represcription. The upper half of the range for price cap carriers would be from 8.48% to 9.52%.[4] Based on the Staff Report and elements of our own analysis, the Coalition recommends the use of a point estimate of 9.00% with a weighted average approach debt-to-equity ratio of 33% debt to 67% equity and a cost of debt of 5.6%. As a threshold matter, the Staff Report provides a significant and credible analysis of the cost of capital faced by telecommunications carriers though there are a number of technical points, explained below, with which we take exception. The Staff Report has a broader purpose of addressing the cost of capital of rate of return carriers. Its results are based on analysis of sixteen publicly traded telecommunications carriers, ten of which are price cap carriers that are expected to be eligible to receive support on the basis of the Connect America Cost Model. Our analysis herein focuses on only the ten price cap carriers included in the Staff Report. Upper and Lower Bounds The table below summarizes data from Appendix K of the Staff Report which displays the results of the upper and lower bound analysis developed by the Staff for two cost models, the Capital Asset Pricing Model (CAPM) and the single-stage Discounted Cash Flow (DCF) model. Arithmetic Average vs. Weighted Averages The Staff Report develops its summary conclusions on the basis of simple arithmetic averages of the results for the companies included in the analysis. While there are various theoretical arguments for using 7

10 arithmetic averages for forecasting prospective expected or required rates of return[5] we must consider as asked by Staff, if such results would be a reasonable reflection of the cost of capital and capital structure of an efficient provider? As should be expected, the ten companies in the analysis vary significantly in size based on capitalization.[6] The Staff Report Appendix K upper and lower bound CAPM and DCF results vary based on how they are averaged, though the differences are not large. The Staff Report s use of arithmetic averages gives disproportionate emphasis to firms with very high levels of lower quality debt. A capitalization-weighted approach can be used to increase the emphasis on investment grade carriers. The midpoint of the range of results using arithmetic or weighted averages remains constant at 8.48%. However, whether components are averaged arithmetically or on a weighted basis have significant effects on the ratio of debt to equity, the cost of debt, the CAPM cost of equity and the weighted average cost of capital. The individual company values displayed below are from the Staff Report.[7] Multiple Modeling Methods As is seen from the Staff Report results for CAPM and DCF models, the results for cost of equity vary by model. Use of multiple models provides an opportunity to observe a zone of reasonableness within which a just and reasonable cost of capital may be selected. The Staff Report properly rejects the singlemodel approach suggested by the American Cable Association ( ACA ).[8] The Coalition offers a third approach to validating a reasonable range for the cost of equity for the Telephone Communications, Except Radiotelephone group of firms.[9] This simple model, which is a variant on CAPM, provides a high level estimate for members of the group, which includes all sixteen companies used in the Staff Report. The Build-Up Model simply sums the risk free rate, the market equity risk premium and an industry-specific risk premium to estimate the cost of equity for the group. The selection of the risk free rate and the market equity risk premium[10] will be discussed in subsequent sections. The risk premium for each industry is developed by Ibbotson using a standardized process.[11] The results of the Build-Up Model are displayed below and present a range of outcomes that recognize the current historically low interest rates and the Commission s infrequency of represcription. The results are highly consistent with the range of outcomes by company provided in the Staff Report. Selection of Risk Free Rate CAPM and the Build-up model each begin their assessment of the cost of equity with a risk-free rate. ACA suggests using a 2.00% 10-year treasury rate as the risk free rate and the Staff Report relies on a 1.92% rate from March 26, 2013.[12] We believe these choices are problematic for three reasons. First, the 10-year rate does not match the economic lives of many of the modeled assets. For example, economic lives for cable investments of all types in the Connect America Cost Model are 20 years or more. Second, use of a rate in the 2% range fails to acknowledge that interest rates are at historic lows and are forecasted to increase substantially over the next several years a period over which the national broadband infrastructure will be built out. The Survey of Professional Forecasters for the first quarter of 2013 published by the Research Department of the Federal Reserve Bank of Philadelphia indicates that over the 2013 to year period, 10-year U.S. Treasury bond returns may average 3.70%.[13] Adding the average spread of 56 basis points between 20-year and 10-year bonds over the past 20 years[14] suggests that the correct risk free rate input to the current modeling is approximately 4.26%. This risk free rate is less than the 87-year average of 5.1% reported by Ibbotson.[15] Third, ACA improperly mixes the use of 10-year bond returns with the Ibbotson Equity Risk Premia of 6.70% which was developed by Morningstar using 20-year Long-Term government bond income returns as the risk free rate benchmark. Selection of Equity Risk Premium The Staff Report acknowledges the regularly published Morningstar Ibbotson equity risk premium as a commonly used source for the long run average historical market risk premium and stated a value of 6.7% as reported in the Morningstar Ibbotson 2011 Classic Yearbook.[16] The Staff Report complained of a lack of access to the underlying data to provide a confidence interval around the reported 8

11 estimated means. Morningstar has recently published results through the end of 2012 and the Long- Horizon Equity Risk Premia, 1926 through 2012 has averaged 6.7%.[17] We believe that the Ibbotson data is appropriate to use and it is readily available in published works for minimal cost. Our only caution is that the risk premium used in any cost of equity analysis should be developed consistent with the risk free rate used in the analysis. That is, if the risk premium used is developed by comparing market returns to 20-year government bond income returns, then the risk free rate used in the CAPM or Build-Up Model should be that of the 20-year government bond income return. To do otherwise would run a substantial risk of over- or under-estimating the cost of equity. Selection of Debt and Equity Weights The Coalition agrees with the Staff Report recommendation to use market value capital structure where the amount of equity is based on the value of each company s stock and debt is included at book value.[18] In addition to the Staff Report reasoning on this matter we note that other key inputs to the determination of the cost of equity, such as the risk premium and beta are developed by data suppliers based on market prices and weights. On the other hand with respect to debt, we note that market values of all debt issues for a carrier may be hard to come by in that some company debt issues are private placements and are therefore not registered or publicly traded. We also note that the averaging method whether simple averages or weighted averages of proxy firms capital structures vary significantly. In this case the simple arithmetic average of debt and equity weights result in a mix of 59% debt and 41% equity as opposed to a market-weighted average mix of approximately 33% debt and 67% equity. Selection of Cost of Debt The Coalition generally agrees with the conclusions reached in the Staff Report that the cost of debt should be based upon values and interest costs reported by the companies in their annual reports to the SEC. The Staff Report suggested using a method that corrects an apparent error in the Commission s rules for calculating the embedded cost of debt but also left the door open to using a current cost of debt calculation and possibly focusing consideration of the cost of debt on firms that have either investmentgrade bond ratings, or times-interest-earned ratios roughly equal to the ratios of firms that have such a rating. [19] The Staff report preliminarily recommended the use of a 6.19% cost of debt based on the arithmetic average of the cost of debt of all 16 companies it evaluated. As noted in the discussion on averaging, the choice to weight the cost results on the basis of size has a significant effect on the results. When limited to the cost of debt for the 10 price cap firms in the Staff Report, the arithmetic average debt cost reported by Staff is 7.00%. When weighted by the amounts of outstanding debt, the value drops to 5.82%. Using weighted values would tend to emphasize the issuers of higher quality debt. The Staff Report made one significant error in its development of the cost of debt. Appendix E reports that the embedded rate was developed from 2012 interest expense divided by the average of outstanding non-current long-term debt at the end of 2011 and This approach understates the total amount of debt and overstates the cost of debt by excluding the current portion of long-term debt on which the carriers continue to pay interest. In the alternative and in order to capture a more forward-looking cost of debt, the Coalition suggests that individual company financial reports (i.e., SEC form 10K) could be used to develop the carrying cost of debt as of the end of 2012 by dividing reported long-term debt interest payment obligations for 2013 by total outstanding long-term debt as of December 31, The results of our analysis follow. Revised CAPM Results We have considered modifications to the Staff Report CAPM modeling that includes debt to equity ratios premised on the inclusion of the current portion of long term debt, revision of the computation of the cost of debt to be based on the forward-looking view of 2013 debt costs reported by the price cap carriers and modifications to the computed cost of equity based on replacement of the single Risk Free Rate input with a current and a forecasted view and the use of the Morningstar Ibbbotson equity risk premia for results through the end of The results of these modifications are presented below.[20] Qualitative Considerations 9

12 The Coalition notes that currently unserved areas present substantial incremental risk to service providers because of high costs and low density. Without explicit support, many unserved areas might never be served. In addition, because CAF II funding is presently intended to last only five years the incremental risk to the carriers is exacerbated. In our view it would not be unreasonable to recognize this incremental risk in adoption of a somewhat higher cost of capital. Selection of Comparable Firms Our analysis has focused on all 10 Price Cap carriers in the Staff Report. Certainly differing subsets of carriers could be considered. For example, ACA evaluated the cost of capital, albeit with many downwardly biased errors, for only five firms. The Coalition recommends the inclusion of all 10 publicly traded price cap carriers in the development of the cost of capital for use in the Connect America Cost Model. Summary Results If the Staff Report data is limited to the 10 price cap carriers in the Staff s analysis and there are no modifications to the Staff Report values, the zone of reasonableness based on statistical inference for the cost of capital is in the range of 7.43% to 9.52% on a total capitalization weighted basis. Following the Staff s suggestion to limit the adopted value for cost of capital to the upper half of the range of results, the adopted costs of capital for use in the Connect America Cost Model should fall in the range of 8.48% to 9.52%. It should be noted that the Staff Report lower bound and midpoint of the range of reasonableness for the cost of capital will be shifted upward by the dual effects of the correct inclusion of a 20-year risk free rate [21] and the use of the most recent Equity Risk Premium published by Morningstar Ibbotson.[22] Conclusions For purposes of finalizing input values in the Connect America Cost Model, the Bureau should not utilize an assumed cost of capital of eight percent, calculated with a ratio of debt to equity of 45:55, and a cost of debt of 6.19%, when adopting final annual charge factors. Instead, the Coalition believes that the zone of reasonableness for the cost of capital for use in the Connect America Cost Model is well above 8.48% and up to 9.52%. The Coalition recommends the use of a point estimate of 9.00% with a weighted average approach debt-to-equity ratio of 33% debt to 67% equity and a cost of debt of 5.6%.[23] [1] A Wireline Competition Bureau staff report analyzing the cost of capital and a related Public Notice were released on May 16, See Prescribing the Authorized Rate of Return: Analysis of Methods for Establishing Just and Reasonable Rates for Local Exchange Carriers, WC Docket No , Staff Report, DA (Wireline Comp. Bur. rel. May 16, 2013); Wireline Competition Bureau Seeks Comment on Rate of Return Represcription Staff Report, WC Docket No et al., Public Notice, DA (Wireline Comp. Bur. rel. May 16, 2013). [2] The Staff Report evaluated sixteen publicly traded firms. Ten of those companies are price cap carriers that may receive support based on the Connect America Cost Model. Those ten companies are: Alaska Communications Systems, AT&T, Century Link, Cincinnati Bell, Consolidated Communications, FairPoint Communications, Frontier Communications, Hawaiian Telecom, Windstream Corporation, and Verizon. [3] Staff Report, Appendix K. The Staff Report results themselves are based on simple unweighted arithmetic averages of the individual company upper and lower bounds of the Staff Report s CAPM and DCF analyses and produce and unweighted average zone of reasonableness of 7.86% to 9.10% with an upper half range of 8.48% to 9.10%. [4] The unweighted company-by-company upper half of the range of reasonableness results from Appendix K is 8.48% to 9.10%. [5] See e.g., Ibbotson SBBI 2013 Valuation Yearbook, Market Results for Stocks, Bonds, Bills and Inflation, , p

13 [6] Market Capitalization is based on the number of shares outstanding reported on each company s K as of various dates in early 2013 multiplied by the closing market price on December 31, Long term debt includes both the long term portion and current-maturities of long term debt reported in the K reports. [7] Debt Cost is reported in Appendix E. All other values are reported in Appendix I1. The Staff Report estimated cost of equity was developed using an Equity Risk Premium of 5.88% and a risk free rate of 1.92%. [8] See Ex Parte notice filed on May 2, 2013 in WC Docket No [9] As reported by Ibbotson SBBI 2013 Valuation Yearbook, Market Results for Stocks, Bonds, Bills and Inflation, SIC Code [10] Ibbotson develops its Long-Horizon Equity Risk Premia on the basis of the S&P 500 total return minus long-term (20-year) government bond income returns. The value reported below is the simple arithmetic average of each annual risk premium for the period of 1926 through [11] Ibbotson SBBI 2013 Valuation Yearbook, pp The complete list of companies used to calculate each industry risk premia estimate can be found athttp://corporate.morningstar.c... [12] ACA 5/2/13 Ex Parte Appendix and Staff Report, paragraph 64. [13] Survey of Professional Forecasters, February 15, 2013, Table 7. [14] rates/pages/textview.aspx?data=yield [15] Morningstar Ibbotson 2013 SBBI Valuation Yearbook, p. 23 and Table B-7. Based on data presented in Table B-7, we estimate a 95% confidence interval of 5.1% +/- 0.56%. [16] Staff Report at paragraph 71. [17] Morningstar Ibbotson 2013 SBBI Valuation Yearbook, Table A-1. Descriptive statistics of the Long-Horizon Equity Risk Premia, including confidence intervals, can be derived from the time series data presented in the table by using Excel data analysis tools. Based on data in Table A-1 we estimate a 95% confidence interval of 6.7% +/- 4.3%. [18] Staff Report at paragraphs 44, 47. [19] Staff Report, paragraph 47. [20] We use the Morningstar Ibbotson inputs for Equity Risk Premium of 6.7% and 2.41% for the December 31, 2012 Risk Free Rate. The forecasted risk free rate of 4.26% discussed above. These results do not reflect the use of statistical inference as used in the Staff Report to develop a zone of reasonableness. [21] The Staff Report risk free rate was that of a 10-year U.S. Treasury bond at 1.92%. The correct value to consider is the 20-year U.S. Treasury Bond that was at 2.41% as of 12/31/12 or, in consideration of the abnormally low interest rate environment, a forward-looking forecast of 4.26% based on the Survey of Professional Forecasters published by the Federal Reserve Bank of Philadelphia. [22] The Staff Report used an Equity Risk Premium of 5.88%. The most recent Equity Risk Premium estimate from Morningstar Ibbotson is 6.70%. [23] In the alternative, if the arithmetic average approach is selected the overall cost of capital should still be set at 9.00% with a debt-to-equity ratio of 60% debt to 40% equity with a cost of debt of 6.6%. 11

14 Between April 30, 2013, and June 25, 2013, one party supplemented its prior submission on the following topic: DETERMINING THE ANNUALIZED COST OF CAPITAL INVESTMENTS Background Hybrid Cost Proxy Model: The HCPM adopts a straight line equal-life-group method of depreciation, using Gomerpertz-Makeham curves. These standard curves describe generalized mortality patterns and are used to determine the probable frequency of plant mortality. To estimate depreciation expenses, the HCPM uses the projected lives and future net salvage percentages for the asset accounts in Part 32 of the Commission's rules. The HCPM also selects a particular set of Annual Charge Factors (ACFs) based on a methodology that is user adjustable and reflects the sum for the three inputs: depreciation, cost of capital, and maintenance costs. CQBAT: The CQBAT model uses the same approach as the HCPM. It adopts a straight-line equal-lifegroup method with expected mortality curves. The lifetimes are also set by the HCPM's values. The Bureau notes that the CQBAT model as submitted in the record does not make public the calculations used to set particular input values in the event a lifetime changes. Question(s) 1) Some of the HCPM values (which also are used by the CQBAT model) may no longer be appropriate (e.g., a 9-year lifetime for a DSLAM). Which, if any, of the projected lives used in the HCPM are outdated and should be modified? If so, what specific modifications would you recommend, and what is the rationale for such change? What additional evidence, if any, would be needed to justify such changes? Comments Thomas Cohen, counsel for ACA Response of the American Cable Association - Expansion on Previous Comments 3. Equipment Salvage Values ACA Recommendation: Because the CACM uses the low end of project equipment lives, the Commission should use higher salvage rates for estimating the recoverable value of equipment at the end of its useful life. In the recent Cost Model Order regarding CACM design, the Commission affirmed that the model will utilize the same economic useful lives as used by the High Cost Proxy Model ( HCPM ), which was adopted to implement provisions in the Telecommunications Act of 1996.[1] These economic useful lives are the low end of projection life ranges found in the depreciation tables in the order adopting the HCPM.[2] The impact of using the low end of useful life ranges is to shorten the depreciation period for equipment, and therefore to amortize its cost over a shorter period of time, which increases the annual levelized cost to serve. This depreciation table also includes ranges for asset classes future net salvage rate. In addition to using the low end of useful life ranges, the current version of the CACM also uses the low end of salvage rate ranges. For example, the model assumes that digital switches will be worth nothing (0 percent salvage rate) after 11 years, rather than 5 percent (5 percent salvage rate) of their current value. In another example, the model assumes that it will cost 30 percent of current value (-30 percent salvage rate) to remove intrabuilding copper rather than 5 percent of current value (-5 percent salvage rate). If the low ends of useful life ranges are to be used, then one would expect that the salvage rate would be higher than the low end of the salvage rate range. For example, the CACM assumes that digital switches 12

15 will need to be replaced after 11 years, rather than the high end of the useful life range of 13 years. A digital switch replaced in 11 years is more likely to have recoverable value than a digital switch replaced in 13 years. Therefore, the salvage rate for an 11-year-old digital switch should not be the low value of 0 percent but rather the high value of 5 percent. Moreover, even when using the high end of the salvage rate ranges, six of the 19 asset classes have negative future net salvage rates, implying there is an additional cost that must be incurred by the end of these assets lives. As ACA has noted in its comments on the WCB Cost Model Virtual Workshop,[3] it is not clear why the model should include any additional costs at the end of the asset lives, especially given that the modeled operating expenses provide funding for certain repairs and replacements. These negative values increase the depreciation expenses, resulting in higher ACFs and therefore greater annual levelized costs for each of the asset categories with negative salvage rates. For example, in the case of the Pole asset category, if a pole costs $100 and has an economic life of 25 years, the -75 percent future net salvage value means that the model provides $175 in capital recovery to the price-cap LECs for the $100 capital expense. As such, $175 would be depreciated over 25 years, rather than the $100 value of the asset. Consequently, ACA recommends that the future net salvage rates used in the CACM be modified in two ways: (1) for asset classes where the high end of the salvage rate range is positive, the Commission should adopt the high end of the salvage rate range, and (2) for asset classes where the high end of the salvage rate is negative, the FCC should adopt a salvage rate of zero. [1] See Connect America Fund/Universal Service Support, WC Docket Nos , , Report and Order, DA , 35 (Apr. 22, 2013) ( Cost Model Order ) ( Based on our review of the record, we now conclude the model will utilize the same economic lives for assets as specified by the Commission previously when it adopted the HCPM. ). [2] See 1998 Biennial Regulatory Review Review of Depreciation Requirements for Incumbent Local Exchange Carriers, CC Docket No , Report and Order, FCC , Appendix B (Dec. 30, 1999). [3] See Thomas Cohen, Comments on WCB Cost Model Virtual Workshop 2012: Determining the Annualized Cost of Capital Investments, available athttp:// (posted Mar. 7, 2013). 13

16 Between April 30, 2013, and June 25, 2013, the Wireline Competition Bureau posted the following new topics in the Connect America Cost Model Virtual Workshop: FINALIZING INPUTS FOR CONNECT AMERICA COST MODEL COST ESTIMATION MODULE Background The Connect America Cost Model (CAM v3.1.2) has two input collections for the cost estimation module. The two input collections contain identical default inputs, except one includes values for Annual Charge Factors (ACFs) calculated with a nine percent cost of capital, and the other includes values for ACFs calculated with an eight percent cost of capital. Parties who have signed the Third Supplemental Protective Order will be able to view the input collections by accessing the model and viewing the ICCQA CAM312ACF8SBI6VoiceCblVoiceFW2 and ICCQA CAM312ACF9SBI6VoiceCblVoiceFW2 ZIP files on the Posted Data Sets page under Model Inputs. The input collections include values for such variables as plant mix, network sizing and sharing, company size categories, operating expenses, capital investments by density and terrain, state property tax factors, regional cost adjustments, bandwidth, business and residential take rate, and state sales tax. Question(s) 1) The Bureau seeks comment on adopting the non-acf default input values currently used in the cost estimation module input collections of CAM v3.1.2 for the final version of CAM. We are separately seeking comment on whether to adopt ACFs that assume an eight percent cost of capital or ACFs that assume a nine percent cost of capital in a follow-up question to the Rate of Return virtual workshop topic. To the extent commenters argue that different inputs should be used, they should describe in detail their proposals and supply specific input values. 2) Given the fraction of costs driven by labor, commodities and electronics along with the expected changes in prices for those inputs, net of productivity gains, Bureau staff believe that it is reasonable to assume static input values in estimating costs in the CAM. To the extent parties disagree, they should specify what assumptions we should make and provide evidence on historical or expected price movements for the costs of labor, fiber, electronics, poles, conduit, and land used in network deployment to support their arguments. Comments Thomas Cohen, counsel for ACA 1. Starting Year Capital Equipment Price Benchmarks ACA Recommendation: Because starting year capital equipment prices provided for the cost model by incumbent local exchange carriers were determined based on data collected in 2011, they do not reflect the decline in market prices. Accordingly, starting year prices for capital equipment should be reduced. The CACM estimates capital equipment costs via a detailed set of tables outlining prices for various components of the network. The data in the tables are drawn from a series of price surveys of members of the United States Telecom Association ( USTelecom ), which for the non-fiber-to-the-premises ( FTTP ) data points was conducted between March and May of 2011 and for FTTP data points in the fall of Prices for communications equipment consistently fall year-over-year. The Board of Governors of the Federal Reserve records the following price declines for equipment relevant to a wireline broadband network:[1] 14

17 Enterprise and home voice equipment: 8.3 percent annual price decline from ; 7.8 percent annual price decline from Transmission, local loop, and legacy central office equipment: 9.3 percent annual price decline from ; 8.7 percent annual price decline from Data networking equipment: 12.3 percent annual price decline from ; 13.1 percent annual price decline from The Commission currently is planning to begin to distribute Connect America Fund ( CAF ) Phase II in 2014, more than two years after the dates of the original price surveys by USTelecom. Thus, for the capital equipment input to be up-to-date, the FCC needs to account for at least two years of price declines from the original capital expenditure estimates. ACA recognizes that it may be preferable to apply a different price deflator to each capital equipment category. However, that level of specificity may not be available. Therefore, ACA recommends, as a balance between expedience and fairness, that the Commission use a single price deflator across all capital equipment price benchmarks, compounded on an annual basis for two years. Based on equipment price data cited above, ACA recommends a 9 percent annual price decline, which is the mid-point of the 4-year and 9-year Compound Annual Growth Rates of the price index of transmission, local loop, and legacy central office equipment. Given the more rapid price declines in FTTP equipment, this price deflator is likely conservative. [1] See Board of Governors of the Federal Reserve System, Industrial Production and Capacity Utilization G.17: Communication Equipment Annual Industry Price Declines, available at (rel. Apr. 19, 2013). 2. Ongoing Capital Equipment Price Adjustment Mechanism ACA Recommendation: Because of the historical deflationary trend in pricing of telecommunications equipment, the CACM should include a mechanism that reduces capital equipment prices over time. The current version of the CACM assumes that prices for capital equipment, once set at inception, will never change, yet, as indicated above prices for telecommunications equipment consistently decline. This assumption leads the CACM to overestimate subsidies as soon as year two of the model, because the use of Gompertz-Makeham survival curves implies that a certain percentage of each asset category will be replaced each year. For longer-lived equipment, such as conduit systems (economic useful life of 50 years), the impact on the subsidies in the model will be small due to the five-year horizon of the CAF and the declining time value of money. However, for shorter-lived equipment, such as digital switches (economic useful life of 11 years), the impact will be significant during the course of the CAF s five-year funding period. Accordingly, to reflect industry cost trends and standard practice among numerous regulators the Commission should add a mechanism that allows capital equipment prices to be reduced over time at a standard rate.[1] [1] LRIC (long-run incremental cost) models adopted by other national regulatory authorities routinely include an automatic price adjusting mechanism so as to model accurately the declining cost of telecommunications capital equipment. The following are examples of this standard practice: The model used by the Australian Competition and Consumer Commission to set prices for five categories of fixed line services includes the functionality to change unit cost trends for capital equipment in the core network and access network. See Model documentation for the Australian Competition and Consumer Commission: Fixed LRIC model user guide Version 2.0, 167 (Analysys Mason 2009) ( The unit cost trends over time can also be defined by the user. ). The model used by Ofcom, the regulator and competition authority for the United Kingdom communications industries, to set middle mile wholesale pricing, includes the functionality to change asset unit costs over time. See Fixed Narrowband Market Review: NGN Cost Modeling Model Documentation v1.0 prepared for: Ofcom, 21 (CSMG Global, 2012). The model used by the Norwegian Post and Telecommunications Authority to set mobile termination rates includes a worksheet titled CostTrends, which allows capital and operational costs to vary over 15

18 time. See Report for NPT: NPT s mobile cost model version 5.1, Model documentation, 3, 5, 29 (Analysys Mason 2009). LRIC models used by ANCOM, the National Authority for Management and Regulations in Communications, to set pricing levels for various fixed line, mobile, interconnection and Ethernet backhaul services, include the capability to change asset prices over time. See Calculation of the costs of efficient provision for some electronic communications services provided at the wholesale level in Romania, Fixed Core Model Documentation, 64 (ANCOM, 2012) ( Price trends are used in order to take into account changes in prices. ). See also Calculation of the costs of efficient provision for some electronic communications services provided at the wholesale level in Romania, Mobile Model Documentation, 54 (ANCOM, 2012); Calculation of the costs of efficient provision for some electronic communications services provided at the wholesale level in Romania, PoI Cost Model Documentation, 19 (ANCOM, 2012); and Calculation of the costs of efficient provision for some electronic communications services provided at the wholesale level in Romania, 14 (ANCOM, 2012). The model used by British Telecommunications plc to set the retail price of call termination and origination for wholesale fixed voice includes the ability to change some asset prices over time. See Long Run Incremental Cost Model: Relationships & Parameters, 129 (BT, 2012) ( Equipment Unit costs are indexed forward using the ASU cost trend. ). Robert Mayer, United States Telecom Association on behalf of ABC Coalition Query 1 Response: The Coalition supports the use of the non-acf default input values currently used in the cost estimation module input collections of CAM v3.1.2 for the final version of CAM. A general description of these non- ACF input values are found in 11 input tables described in Appendix 6 of the CAM Model Methodology document revised 5/22/2013, whereas the actual input values are found in the posted data sets mentioned above. While the Coalition recommends no further changes to these input values, at least one commentator has offered suggested changes that the Coalition views as inappropriate. Specifically, ACA responded on May 21, 2013, to this virtual workshop topic proposing that all capital equipment price benchmarks be reduced by 9% per year for two years to reflect the passage of time between the data collection populating the material and installation costs in 2011 and the use of the CAM results at the beginning of This commentary by ACA is consistent with its ex parte filed on May 17, Apparently lacking a more nuanced set of price information for every capital equipment category, ACA viewed the use of a single price deflator across all capital equipment price benchmarks as a balance between expediency and fairness. The ACA proposed adjustments are not reasonable. First, as the Bureau notes, if a party argues that different inputs should be used, they should describe in detail their proposals and supply specific input values. The Capex input table provides the material and installation costs for the plant build developed in CAM. However, ACA s proposal refers generically to capital equipment costs via a detailed set of tables outlining prices for various components of the network, which apparently refers to the Capex input table. While it seems apparent that ACA is not referring to installation costs, the Coalition still must guess what specific material input values in the Capex input table that ACA argues should be updated. As the Commission well knows, given the large number of inputs needed for CAM, the input determination process takes a significant amount of time before the Commission is able to determine reasonable cost results, as was also the case in the determination of inputs for the synthesis cost model made in CC Docket No In addition, material costs for facilities and installation costs are generally more significant contributors to the costs developed by CAM than material costs for equipment. If the Commission were to consider updating material and installation costs used by CAM, all material and installation costs, and not just some cherry-picked subset, should be examined. Second, although ACA discusses updating all telecommunications equipment prices, it asserts that it may be desirable to categorize equipment in some undefined groupings, but ACA does not find it expedient or 16

19 fair to do so. Instead, ACA argues that all telecommunications equipment prices should be reduced by a common percentage. However, it is not apparent that prices for all communications equipment found in CAM, e.g., cabinets, batteries, network routers and optical equipment, would change at approximately the same rate over 2012 and Consequently, the Coalition views ACA proposal as exhibiting more expediency than accuracy. Third, ACA proposes to reduce equipment prices by 9% per year compounded for two years, i.e., a cumulative reduction of 17.2%. ACA describes this annual percentage decline as the mid-point of the 4- year and 9-year Compound Annual Growth Rates of the price index of transmission, local loop, and legacy central office equipment developed by the staff of the Federal Reserve Board. This price index covers prices for fiber optic long-haul and local loop equipment. The attached chart shows the percentage declines in fiber equipment from 2003 to 2011, which the Coalition presumes is the information relied upon by ACA. In addition, this chart shows the trend line for the percentage price changes. Clearly, the percentage price declines for fiber equipment have been getting smaller over time. Based on this data, it is not reasonable to take averages, as apparently ACA has done, since such averages would project percentage price declines for 2012 and 2013 substantially larger than the strong trend implies. Using the trend line, a more reasonable estimate would be a decline of 6.4% for 2012 and a decline of 6.0% for 2013 or a 2-year decline of 12.0%. Thus, if the Commission wished to update fiber long-haul and loop equipment prices from the end of 2011 to the end of 2013, decreasing such material prices by 12% would be more appropriate than the rate recommended by ACA. Now, the material prices found in the Capex input table include prices for more than just equipment needed for a FTTH build. Specifically, the Capex input table also includes material prices for communications structures fiber cable of various types, poles, conduit, land and buildings. The only public price index known by the Coalition related to the material prices for structure found in CAM is the price index for private fixed investment in communications structures found in Table of the U.S. National Income and Product Accounts published by the Bureau of Economic Analysis (BEA). Communications structures in this table include cable, poles, conduit and buildings. Over 2003 to 2011, this price index for communications structures has increased an average of 5.3% per year, although the average for the most current four years is 3.8%. The mid-point of the 4-year and 9-year averages is an annual increase of 4.5%. As can be seen in the diagram below, while there are significant year-to-year variations, the trend line is nearly flat. Clearly, the fiber-based price index recommended by ACA is not reasonable for communications structure. If Commission were to decide to update material prices for fiber cable, poles, conduit and buildings, a 2-year upward adjustment of 6.9% would be consistent with recent experience of the annual percentage change in BEA s communications structure price index. There are a number of candidates to use to update installation costs, which are based on numerous labor rates. The Consumer Price Index (CPI) published by the Bureau of Labor Statistics (BLS) measures inflation as experienced by consumers in their day-to-day living expenses and has often been used in private sector collective bargaining agreements to modify over the term of a contract the wages to unionized employees. The following chart shows the annual percentage changes in the CPI for all urban consumers from January 2003 to April 2013, which is the most currently available data published by the BLS. The average annual percentage change in CPI-U for 2012 was 2.1% and the average annual percentage change in CPI-U for 2013 based on the first 4 months of 2013 was 1.5%. Using these annual percentage changes, a 2-year adjustment increasing labor rates by 3.5% would be consistent with the CPI-U data. Clearly, CPI-U does not directly measure inflation for total labor compensation or even just wages. The BLS conducts the National Compensation Survey, which provides employer costs for employee compensation. Using this survey, the BLS the shows the employers average hourly cost for total labor compensation including break outs by wages and salaries, and total benefits. In addition, BLS breaks out data by major industrial groups based on the North American Industry Classification System (NAICS). The Information sector, which includes publishing, motion pictures, sound recording, broadcasting and data processing in addition to telecommunications, is the lowest level of industry detail that includes ILECs. The next chart shows annual percentage change in total compensation per hour worked for the Information sector from 2004 to The average annual percentage change for 2012 was 5.8%. Assuming that the 2013 percentage will be the same, the 2-year average adjustment would increase labor rates by 12%. 17

20 While the BLS data are imperfect matches, there exists another data source, published by the BEA, that is a much better match. The BEA regional economic accounts provide statistics about employment and compensation for the telecommunications industry (NAICS industry 517). Using the BEA data, the average annual compensation per telecommunications job can be calculated. The percentage change in annual compensation per job is shown in the diagram below. The average annual percentage change for 2012 was 3.8%. Assuming that the 2013 percentage will be the same, the 2-year average adjustment would increase labor rates by 7.7%. If Commission were to decide to update installation costs by two years, then an increase of labor rates by 7.7% would be a reasonable estimate. While the Coalition has provided some alternatives that improve on the suggestions made by ACA regarding updating material and installation costs, the Coalition still believes a less speculative approach is more appropriate, which is not to make any inflation adjustment to the material and installation costs found in CAM. Query 2 Response: The Coalition agrees with the Staff s presumption to keep the non-acf default input values static for purposes of developing final costs. To do otherwise would likely render the model-derived cost estimates unrealistic. Further, while making inflation adjustments to material and installation costs are feasible, CAM is ill-suited to incorporate any impact of material and installation cost changes over time. Also, it would be unreasonable and impractical to limit changes to only material and installation costs, since other economic valuation components may also change over time. For example, operating expenses are also likely to change over time. ACA also contends in its virtual workshop comments that CAM should be modified to include a mechanism that reduces capital equipment prices over time. But, as discussed above, equipment prices are not the only material and installation costs that change over time. In fact, equipment prices are less significant contributors to the costs developed by CAM than other material and installation costs. Again, ACA s specific suggestion is neither practicable nor reasonable. ACA also points to national regulatory bodies of other nations to support its contention that the capability to reflect a continuing inflation adjustment is warranted. Given that it is not reasonable to reflect a continuing inflation adjustment, it would be a waste of valuable resources to add a capability that should not be used. Next, ACA appears to believe that incorporating a mechanism that reduces capital equipment prices over time is necessary because any capital investment that replaces capital that has served its economic life needs to reflect the future price at the time of replacement. But, there is no replacement cost calculation in the current, greenfield version of CAM. Replacement cost calculations only occurred in version of the brownfield model, which was not adopted by the Commission. Finally, ACA argued in its May 17, 2013 ex parte, but not in its virtual workshop comments, that because CAM uses the low end of project equipment lives, the Commission should use higher salvage rates for estimating the recoverable value of equipment at the end of its useful life. This recommendation should be rejected. ACA s ex parte comments on salvage values explicitly discussed lives and salvage values for digital switches and copper cable. Since CAM does not develop any costs for digital switching or copper cable, these comments are irrelevant and should be disregarded. ACA expressed incredulity that any asset can incur additional costs at the end of its life. The Coalition is baffled that ACA is unaware that cost of removal occurs at the end of the useful life of an asset, and a net salvage percentage reflects the salvage value of asset less the cost of removing the asset from the network so as to be able to receive that value. The use of negative net salvage percentages for some asset categories is reasonable and requires no change to the salvage values used in CAM. Robin Tuttle, counsel for ACS Question 1 18

21 ACS has repeatedly documented the limitations of the Connect America Cost Model ( Model ), specifically that the Model does not offer any accommodations to address the unique difficulties and associated high costs of serving insular price cap local exchange carrier ( LEC ) areas.[1] While ACS submitted for review by the Bureau its own model of the satellite, microwave, and undersea cable transport costs inherent in serving Alaska,[2] ACS has continued working on a proposal of adjustments to the Model that will address the unique demands and costs of providing voice and broadband service in Alaska. ACS expects to submit these proposed adjustments in the coming weeks. In the meantime, ACS urges the Bureau to consider a number of varying factors that significantly raise the costs to provide service in Alaska. Higher costs in Alaska are not indicative of either inefficiencies or opportunities for cost savings. As a company that is facing more intense competition more pervasively across its service territory than any other price cap carrier, ACS is compelled by market forces to operate efficiently. Costs for providing service in Alaska are driven by the extremes in the state s remote location, challenging geography and climate, and the associated limitations in infrastructure. As a result, network construction, operation, and maintenance practices differ from those that many carriers experience in the Lower 48, and create significantly higher costs. For example, Alaska s climate requires network operators to bury fiber more deeply than carriers do in the Lower 48 states. Engineering standards must account for extreme temperature conditions and travel to and from distant locations adds time and cost to build and operate a network. The state s remote location translates into higher transportation, fuel, and labor costs that are unique to Alaska. Also, building, upgrading, and maintaining the network is restricted by the short summer construction season in Alaska and there are additional costs to mobilize and de-mobilize the construction effort at the beginning and end of abbreviated construction seasons, as well as higher labor costs from paying overtime necessary to achieve deployment objectives during the short season. ACS has well documented in these proceedings how these and other critical inputs represent higher costs and expenses, but also require longer build-out cycles than other carriers typically require.[3] The state s very small population base and low population density means that even an efficient carrier cannot achieve the purchasing power and economies of scale available to industry giants in the Lower 48 states.[4] The Bureau should not finalize input values for a national Model until it can ensure that the Model s cost module will accurately capture these differences for Alaska. Using national average costs will result in false conclusions about the costs for carriers like ACS to provide broadband, impacting their ability to comply with the Commission s CAF Phase II broadband mandate. ACS is not able to gain the same efficiencies of scale or scope as larger price cap carriers serving more than a single state. ACS cannot average its network deployment costs across multiple states with varying cost-causative characteristics. Unlike nationwide carriers, such as AT&T and Verizon, that enjoy significant economies of scale and serve areas with widely varying costs, ACS is serving only the sparsely populated, high-cost state of Alaska. ACS cannot simply join together with other carriers to gain purchasing power or other efficiencies. In the first place, its service territory is not contiguous with those of other price cap carriers. Second, carriers design and build their networks differently, limiting the commonality that might facilitate efficiencies. It is unreasonable to expect that ACS s costs could ever be compared to those of a nationwide carrier, or even a regional price cap carrier serving the Lower 48 states, for the purpose of assessing efficiency. Like any other carrier in a competitive environment, ACS seeks to minimize its costs and maximize output, but the fact remains that ACS lacks the economies enjoyed by carriers serving the Lower 48. The Model should account for these differences. Just as importantly, the Model should not assume that an efficient carrier will always be one of the size and scale that is common amongst the largest price cap carriers. The Communication Act provides for high cost support that is sufficient and predictable, but assumptions about a provider that do not exist (such as a large price cap carrier in Alaska) will never lead to high cost support that meets the legal standards. Indeed, on the OpEx side, the Model explicitly acknowledges differences among the expense profiles of efficient carriers, providing for seven size classifications of carriers, including extra-extra small, extra small, small, medium, and large companies segregated by urban/suburban and rural density areas. Yet even here, by classifying ACS as a medium carrier, the model falls short of capturing ACS s true 19

22 circumstances. First, the Model defines medium carriers as having between 100,000 and 1 million lines. ACS, with just over 100,000 lines, falls at the very bottom of that range and, as a single-state carrier operating in an extremely challenging service area, has far more in common with carriers in the small category, defined as those with between 4,000 and 99,999 lines. Further, as ACS, like most carriers, continues to experience line loss, it is possible that it will fall below the 100,000-line mark before the end of the CAF Phase II build-out period. Second, even with respect to medium carriers, based on industry data, the Model incorporates a substantial negative adjustment to the operating cost profile of carriers classified as medium, suggesting that they are far more operationally efficient than large carriers. This suggestion stands in marked contrast to the substantial positive adjustments for the three categories of small carriers. The Model s hypothesis that medium carriers enjoy the greatest operational efficiency among all sizes of carrier is counterintuitive, and based on statistical calculations of limited validity. Specifically, there are only seven carriers within the medium range, and the NECA investment and expense data for those carriers, on which the Model calculation is based, vary markedly among the seven. On balance, the average investment-to-expense ratio in the raw data is much closer to that experienced by larger carriers than the results of the CostQuest data analysis ultimately reflect. It is also important to note that the expense factors employed in the CostQuest analysis represent an average of 2008, 2009, and In 2008 AT&T s operating companies in Kentucky, Arkansas, Nevada and Kansas were considered medium companies in CostQuest s analysis, but they were dropped from the subsequent data periods. Thus, for two out of the three years analyzed, the medium company group consisted of only ACS, Cincinnati Bell, Hawaiian Tel, Consolidated, Rock Hill, TDS, and PRTC. This results in there being 60 percent fewer loops and 50 percent less total plant in service in 2009 and 2010 data periods. Given the small, uneven sample size of the set of medium carriers, the wide range of expense ratios those data reflect, and the counterintuitive nature of the downward adjustment currently included in the Model, ACS believes that there is little empirical support for assuming that, for the same level of investment, a medium size carrier would have operating expense for cable a wire investment that is so substantially lower than that experienced by much larger companies. The only justification would be to accept the hypothesis that medium carriers are inherently more efficient than larger carriers. Further, if the model accepts that efficient companies may have varying levels of operating expense due to size, if should also accept that different sized companies may have different levels of capital expense levels as well. Yet, the Bureau has indicated that an efficient carrier should have the same level of capex as the largest carriers when clearly a carrier such as ACS would pay more for the same piece of equipment than would AT&T and Verizon. As a result, ACS believes that the OpEx adjustment for medium carriers should be set to zero. This would reflect the fact that there are few medium carrier data points on which to make an assessment, and the fact that those data are relatively similar to the large carrier data taken as a whole. Alternatively, for the reasons stated above, ACS believes that the Commission should reclassify ACS as a small carrier for purposes of the Model. ACS, as a carrier that serves a challenging service area within a single state, and a shrinking line count that barely exceeds 100,000 today, has far more in common with the carriers classified as small for purposes of the Model, and should be classified as such. [1] See Connect America Fund; High-Cost Universal Service Support, Comments of Alaska Communications Systems, WC Docket Nos and (filed Feb. 27, 2013). [2] See Letter to Marlene H. Dortch, Secretary, Federal Communications Commission, from Karen Brinkmann, Counsel for Alaska Communications Systems, Request for Connect America Fund Cost Models, Public Notice in WC Docket Nos and , DA (Wireline Competition Bur., rel. Dec. 15, 2011), Submitted Pursuant to Second Protective Order in WC Docket Nos and , DA (Wireline Competition Bur., rel. Feb. 10, 2012), submitting the ACS model. [3] See Connect America Fund; High-Cost Universal Service Support, Comments of Alaska Communications Systems Group, Inc., WC Docket Nos and (filed Feb. 1, 2012); Connect America Fund; High-Cost Universal Service Support, Comments of Alaska Communications Systems Group, Inc., WC Docket Nos and (filed July 9, 2012); Connect America Fund; High-Cost Universal Service Support, Reply Comments of Alaska Communications Systems Group, Inc., WC Docket Nos and (filed July 23, 2012); Letter (Ex Parte Notice) to Marlene H. Dortch, Secretary, Federal Communications Commission, from Karen Brinkmann, Counsel to Alaska 20

23 Communications Systems Group, Inc., Developing a Unified Intercarrier Compensation Regime, et al., CC Docket Nos and 96-45, WC Docket Nos , , , and 10-90, WT Docket No , and GN Docket No (filed April 27, 2012); Letter to Marlene H. Dortch, Secretary, Federal Communications Commission, from Karen Brinkmann, Counsel to Alaska Communications Systems Group, Inc., Developing a Unified Intercarrier Compensation Regime, et al., CC Docket Nos and 96-45, WC Docket Nos , , , and 10-90, WT Docket No , and GN Docket No (filed May 11, 2012), submitted subject to Second Supplemental Protective Order in WC Docket Nos and 10-90; Letter (Ex Parte Notice) to Marlene H. Dortch, Secretary, Federal Communications Commission, from Richard Cameron, Assistant Vice President and Senior Counsel for Alaska Communications, Developing a Unified Intercarrier Compensation Regime, et al., CC Docket Nos and 96-45, WC Docket Nos , , , and 10-90, WT Docket No , and GN Docket No (filed July 27, 2012); Letter (Ex Parte Notice) to Marlene H. Dortch, Secretary, Federal Communications Commission, from Richard Cameron, Assistant Vice President and Senior Counsel for Alaska Communications, Developing a Unified Intercarrier Compensation Regime, et al., CC Docket Nos and 96-45, WC Docket Nos , , , and 10-90, WT Docket No , and GN Docket No (filed August 28, 2012); Alaska Communications CAF II Model, FCC Workshop, Sept , 2012, presentation by David Blessing, Karen Brinkmann, and Richard Cameron, available athttp://transition.fcc.gov/wcb/... FCC Connect America Phase II Cost Model Virtual Workshop, Comments by Alaska Communications Systems, [4] The State of Alaska has a population density of 1.2 persons per square mile, the lowest in the nation, compared with 87.4 for the United States as a whole. The state s largest population center, Anchorage, ranks 63rd on the list of the nation s largest cities with a population of 298,610 and a population density of persons per square mile. The two cities above and below it in population on the Census Bureau s list, Lexington, Kentucky and Stockton, California, have population densities of 1,042.8 and 4,730.1 persons per square mile respectively. See Question 2 ACS agrees with the Bureau that it is reasonable to use static input values when estimating costs in the Model. Some input costs may increase while others may decrease over time, but the accuracy of the Model will be diminished if changes are made to only some input values without conducting a thorough review of changes that might be appropriate for all input values and any such predictions would be speculative at best. Moreover, such a review would be time consuming and cause delay in completing the Model and implementing CAF Phase II support. The best course is to use static input values for the covered period. 21

24 CONNECT AMERICA FUND-INTERCARRIER COMPENSATION RECOVERY MECHANISM SET ASIDE AMOUNT Background In the USF/ICC Transformation Order, the Commission established an annual funding target of $4.5 billion for high-cost universal service support. Within the $4.5 billion budget, the Commission set aside up to $1.8 billion annually for a five-year period to support areas served by price cap carriers. This amount includes the support that price cap carriers receive through the Connect America Fund intercarrier compensation (CAF-ICC) recovery mechanism. The CAF-ICC recovery mechanism is an explicit support mechanism that replaces the implicit support previously received by carriers from carrier-to-carrier revenues. Question(s) 1) In order to finalize the cost model and identify the census blocks that will be funded, the Bureau needs to specify the amount of funding to be allocated among census blocks through the cost model or competitive bidding. The Bureau forecasts that over a five-year period, from 2015 to 2019, price cap carriers will draw an average of roughly $50 million per year of support from the CAF-ICC recovery mechanism. If the Bureau were to set aside $50 million from the $1.8 billion price cap carrier budget when finalizing the model, this would mean that $1.75 billion in support would be distributed through the model or competitive bidding. Is it reasonable to utilize a straight average when forecasting the price cap carrier draw from the CAF-ICC recovery mechanism? Is $50 million a reasonable amount of support to set aside for the CAF-ICC recovery mechanism in price cap areas? We encourage the price cap carriers to submit their current projections of their anticipated CAF-ICC draw over the relevant time period. Parties that argue that a different methodology should be used should describe in detail their proposals and identify all underlying assumptions for a specific set aside amount for the CAF-ICC recovery mechanism. Comments Robert Mayer, United States Telecom Association on behalf of ABC Coalition The Coalition believes that it is reasonable to set aside $50 million from the $1.8 billion in support to recognize the average draw of approximately $50 million per year of support from the CAF-ICC recovery mechanism. Additionally, the Coalition believes using a straight average when forecasting the price cap carrier draw from the CAF-ICC recovery mechanism, resulting in $1.75 billion of support to be distributed through the model or competitive bidding. 22

25 SUPPORT THRESHOLDS Background In the USF/ICC Transformation Order, the Commission adopted a methodology that will target support to areas that exceed a specified cost benchmark, but not provide support for areas that exceed an 'extremely high cost' threshold. With regard to the support benchmark, the Commission stated that it would use the model to identify those census blocks where the cost of service is likely to be higher than can be supported through reasonable end-user rates alone. With regard to the extremely high cost threshold, the Commission also concluded that "a small number of extremely high-cost census blocks that should receive funding specifically set aside for remote and extremely high-cost areas... rather than receiving CAF Phase II support." The Commission anticipated that no more than 1 percent of all American household would be in such remote and extremely high-cost areas. Finally, the Commission directed that "[t]he threshold should be set to maintain total support in price cap areas within our up to $1.8 billion annual budget. In the Model Design PN, the Bureau sought comment on how to set the funding and extremely high-cost thresholds. It specifically sought comment on whether the Bureau should first determine the funding threshold and then use the budget to determine the extremely high-cost threshold, or if it should first determine the extremely high-cost threshold and then use the budget to determine the funding threshold. Both ACA and NASUCA urged the Bureau to use the former approach, and set the funding threshold first. Question(s) 1) One possible method for establishing the support threshold would be to estimate the average revenue per user (ARPU) that could be reasonably expected from voice and broadband services and make adjustments to take into account that not all locations passed will necessarily subscribe to one or both services over the full term of Phase II support. Is this an appropriate way to set the support threshold? 2) The Bureau recognizes that there may be different take rates for standalone voice service, standalone broadband service, and a package that includes both voice and broadband, and that the number of locations connected (and therefore able to subscribe) will increase over time as deployment progresses. The Bureau previously sought comment (Calculating Average Per-Unit Costs) on the assumption that, on average, 80% of locations would subscribe over the Phase II time horizon, noting that take rate has a small impact on the cost per location passed. (To illustrate the point, if 60% of locations subscribe at the beginning of Phase II and 100% subscribe at the end of Phase II, that would represent an average subscription rate of 80% over the fiveyear period.) What assumptions for ARPU and take rate are appropriate for purposes of setting the funding threshold? 3) The table below shows the support threshold for various take rate-arpu combinations. Would adopting a funding benchmark in the $40 to $50 range be a reasonable approach? To the extent commenters believe the funding threshold should set higher or lower, they should identify with specificity their underlying assumptions about ARPU and take rate. Average Take Rate over Phase II ARPU 50% 60% 70% 80% $50 $25 $30 $35 $40 $60 $30 $36 $42 $48 23

26 $70 $35 $42 $49 $56 $80 $40 $48 $56 $64 4) Given the Phase II budget of up to $1.8 billion, adopting a support benchmark in the $40 to $50 range could result in an extremely high-cost threshold between $145 and $155 per location passed, under version of the Connect America Cost Model with default input values. Is this a reasonable range for the extremely high-cost benchmark? 5) Are there other methods of calculating the support threshold for Connect America Phase II support? For instance, would basing the funding benchmark on a specified cost percentile, such as the 95th percentile, be appropriate? Are there other methods that the Bureau should consider? Comments Thomas Cohen, counsel for ACA Q1. One possible method for establishing the support threshold would be to estimate the average revenue per user (ARPU) that could be reasonably expected from voice and broadband services and make adjustments to take into account that not all locations passed will necessarily subscribe to one or both services over the full term of Phase II support. Is this an appropriate way to set the support threshold? The ACA agrees that the support threshold should be set by estimating the ARPU that could be reasonably expected from voice and broadband services and then making an adjustment to take into account that not all locations passed will necessarily subscribe to one or both services over the expected funding period. Put another way, the support threshold should be equivalent to the expected ARPU from voice and broadband services multiplied by the expected take rate for those services. Q2. The Bureau recognizes that there may be different take rates for standalone voice service, standalone broadband service, and a package that includes both voice and broadband, and that the number of locations connected (and therefore able to subscribe) will increase over time as deployment progresses. The Bureau previously sought comment (Calculating Average Per-Unit Costs) on the assumption that, on average, 80% of locations would subscribe over the Phase II time horizon, noting that take rate has a small impact on the cost per location passed. (To illustrate the point, if 60% of locations subscribe at the beginning of Phase II and 100% subscribe at the end of Phase II, that would represent an average subscription rate of 80% over the five-year period.) What assumptions for ARPU and take rate are appropriate for purposes of setting the funding threshold? The take rate used to set the funding threshold should be the same take rate used to develop cost model assumptions. The core principle that should guide the adoption of a take rate for setting the funding threshold is that the rate used should be the same take rate used to develop cost model assumptions. The FCC should not use one take rate for estimating costs and a different take rate for estimating expected revenues. Doing so would not only contradict widely accepted principles of network-planning and business case modeling, it would also over-compensate operators receiving Connect America Fund (CAF) Phase II funds. When rational operators are making network investments, they use the same take rate to estimate both costs and expected revenues. From the perspective of an operator making a network investment, it makes no sense to use two different take rates the take rate is a common assumption used across cost and revenue models to determine with some level of precision whether a network build-out will produce a positive return. Using a different (and by definition, inaccurate) take rate for the cost model and for the revenue model will lead to an operator either investing in a network build-out with a negative return or passing on a build-out that would have resulted in a positive return. Any investor relying on a business 24

27 case with different take rates will have an inaccurate view of the operator s appropriate cost of capital. Using different take rates for costs and revenues ultimately leads to inefficient investment. In a typical network investment, the fixed cost of the network up to the curb of potential subscribers is incurred upfront, while the variable ( success-based ) costs of a drop, network interface device and customer premise equipment are not incurred until locations subscribe to the operator s service. This reduces the risk that operators will invest in infrastructure that will go unused, and reduces the present cost of the network investment because the value of money declines over time. The CAF model, however, compensates operators for these variable costs from the beginning of the Phase II funding term. The use of a levelized costing mechanism amortizes all costs, including variable costs, over the entire term of the CAF funding. In effect, the CAF model compensates operators for variable costs years before they happen. Even if the same take rate is used to establish the support threshold as is used to dimension costs, operators will still be over-compensated slightly. If one uses a higher take rate to estimate costs (say, 80%) and a lower take rate to estimate support thresholds (say, 50%), the excess subsidies will be amplified. The CAF model will include the variable costs for building out drops, NIDs and customer premise equipment to 80% of locations, while the setting of the support threshold suggests that only 50% of locations will actually end up subscribing. In this example, the CAF model would model funding for millions of drops, NIDs and customer premise equipment units that will never be installed. The Wireline Competition Bureau appears to suggest, that take rate has a small impact on the cost per location passed. This statement is a reference to an example offered by the WCB in the Virtual Workshop session on Calculating Average Per-Unit Costs ( ACA disagrees. The Calculating Average Per-Unit Costs Virtual Workshop example did not demonstrate that the impact of the take rate on the cost per location passed was immaterial, but rather that it was small relative to its impact when using cost per subscriber to estimate costs. In the example given by the WCB, increasing the take rate from 50% to 70% increased the cost per subscriber by $89, but only increased the cost per location passed by $4. But even a $4 change in the support threshold can have a huge impact on the size of the subsidies distributed and the number of locations supported by the CAF. For example, lowering the target benchmark from $64 to $60 adds an additional 444,308 locations to the CAF that were not previously subsidized. In addition, every dollar by which the target benchmark is lowered adds an additional dollar of subsidy funding to all other locations covered by the CAF. The net result is hundreds of millions of dollars of excess subsidies over the five-year term of the Phase II CAF as well as eliminating wireline broadband subsidies for 100,000 or more locations above the technology cutoff. As the chart below demonstrates, shifting the support threshold ( target benchmark ) by increments of $5 has enormous impact on the number of locations served by the CAF and the number of locations relegated to the Remote Areas Fund.[1] In terms of take rate, a $5 shift in the target benchmark at the levels on the chart is indicative of a shift of less than 10% in the take rate, regardless of what ARPU benchmark is used. That is, if one assumes an $80 ARPU benchmark for voice and data services, the difference in take rate between a $50 target benchmark and a $55 target benchmark will be 6.25%.[2] We will use the shift from a target benchmark of $55 to $50 to illustrate the financial impact of a shift in $5 of the target benchmark (or put another way, the financial impact of a shift of a <10% in the expected take rate). To understand the financial impact of the shift, one needs to consider two components: the excess subsidy given to locations between $55 and the alternate technology cutoff (in this case, $145) and the subsidy given to locations between $50 and $55. The total of these subsidies is equal to the value of subsidies that would have gone to high-cost CAF locations above $145 that are now relegated to the uncertainties of the Remote Areas Fund. The total excess subsidy over the five-year term of CAF Phase II funding is $1,450,651,058 or 16.5% of all CAF Phase II funds.[3] The portion that is excess subsidy given to locations between $55 and the alternate technology cutoff is $1,325,146,800.[4] The excess subsidy given to locations between $50 and $55 over the five-year term of Phase II CAF is $125,504,258.[5] These subsidies not only would provide operators with excess returns, they would also deprive 242,362[6] truly high-cost locations of guaranteed 25

28 wireline broadband. So apparently small changes in the take rate can lead to massive misallocations of CAF funds. The take rate used should be the terminal broadband take rate of the expected funding period, not a blended take rate that averages adoption over the funding period. As noted previously, in a typical network investment, the fixed cost of the network up to the curb of potential subscribers is incurred upfront, while the variable ( success-based ) costs of a drop, network interface device and customer premise equipment are not incurred until locations subscribe to the operator s service. This reduces the risk that operators will invest in infrastructure that will go unused, and reduces the present cost of the network investment because the value of money declines over time. The CAF model, however, compensates operators for these variable costs from the beginning of the CAF Phase II funding term. The use of a levelized costing mechanism amortizes all costs, including variable costs, over the entire term of the CAF funding. CAF recipients are subsidized for all variable successbased investments from the first day CAF funding is disbursed, even though most of the investment will not be incurred until years later. Given the principle that the take rate used to estimate revenues should be the same as the take rate used to estimate costs, it is not reasonable to use a blended average take rate that accounts for escalating adoption over the assumed funding period. Instead, the assumed take rate should be the terminal take rate; that is, the take rate expected at the end of the assumed funding period. If operators are receiving subsidies for 80% adoption from the beginning of the funding period, their support threshold should be set based on the same assumption.[7] The take rate used to estimate costs and revenues should be 90%. When developing cost estimates for the National Broadband Plan, the FCC, in partnership with CostQuest Associates, developed a model for expected broadband adoption. The terminal rate of adoption in this model was nearly 90%.[8] The National Broadband Plan s adoption curve was developed by mapping the Gompertz mathematical model for forecasting technological adoption against broadband adoption data that has been collected by the Pew Internet and American Life Project [9] since The most recent instances of this survey showed broadband adoption among US adults at 65% in December 2012, and 66% in April 2012.[10] Plotting this data point on the FCC s forecast broadband adoption curve suggests that nearly 90% adoption will be reached within six years. Given the five-year term of CAF Phase II, and the fact that funding will not be distributed until 2014,[11] it seems likely that broadband adoption will reach nearly 90% by the end of CAF Phase II. As we previously argued, the broadband take rate used for estimating costs and revenues should be the terminal take rate of the expected funding period. Given the nature of the CAF s funding mechanism and the FCC s own approach for forecasting future broadband adoption, 90% is a reasonable take rate to use to estimate costs and revenues. The ARPU assumption should be based on a weighted average of the ARPU of the minimum broadband and voice services required by the FCC. Price cap carriers largely offer uniform national pricing for DSL broadband and voice services. The presence, or lack thereof, of a competitive provider in a given territory does not typically affect this pricing. Therefore, non-promotional pricing for broadband and voice from any area where 4/1 broadband or greater is available provides useful benchmarks upon which to base the ARPU threshold. Given the unequal distribution of CAF-eligible locations across different operators service areas, the simplest and most equitable way to average these ARPUs is to weight the ARPUs by each price cap carriers share of total CAF-eligible locations.[12] Additionally, while advertised non-promotional pricing for 4/1 broadband and voice service is a reasonable proxy for ARPU, it does not capture the entire ARPU for customers subscribing to both 4/1 broadband and voice services. In the case of packages that do not include unlimited calling,[13] some subscribers will incur additional voice usage charges. So any analysis based on pricing benchmarks will inevitably be conservative. 26

29 Based on ACA s research, the recommended ARPU benchmark should be $71. For this analysis, we first determined the mix of CAF-eligible locations located within each price cap carriers territory using the latest version of the Connect America Cost Model. We then researched the lowest non-promotional, non-contract price advertised for broadband that had at least 4 Mbps downstream and 1 Mbps upstream and voice packages with unlimited local and long-distance minutes, if available. In the cases where pre-packaged bundled offers meeting those requirements were cheaper than a la carte pairings, we used those prices as benchmarks. The output of ACA s research and analysis is below: ACA s recommended support threshold is $64, based on a 90% take rate and $71 ARPU benchmark. Given the following arguments previously set forth: The take rate used to estimate costs should be the same take rate used to estimate revenues, The take rate used for both the cost and revenue estimates should be the terminal take rate, A 90% take rate can reasonably be expected by the end of the funding period, and A conservative ARPU benchmark is $71, the support threshold should be no less than $64. [1] For this analysis, ACA has assumed a total annual fund size of $1.75 billion, changing the alternate technology cutoff to dimension the fund size consistently across varying target benchmarks. [2] ($55/$80) ($50/$80) =.0625 = 6.25%. [3] To determine the excess subsidies, we changed the target benchmark from $50 to $55 but held the number of Remote Areas Fund ( RAF )-eligible locations constant by keeping the effective cost ceiling constant in this case, $145. The difference in the total support-capped funding between the two support scenarios is $24,177,518 a month, or $1,450,651,058 ($24,177,518 * 12 months * 5 years). [4] 4,417,156 locations overlap in the support scenarios addressed in the previous note. All of these locations receive an excess subsidy of $5 a month ($55 - $50) when the target benchmark is lowered to $50. Therefore, the excess subsidy for these locations = 4,417,156 locations * $5 * 12 months * 5 years = $1,325,146,800. [5] The excess subsidy for locations between $50 and $55 is simply the difference between the total excess subsidy ($1,450,651,058) and the excess subsidy given to locations between $50 and $145 ($1,325,146,800). 1,450,651,058-1,325,146,800 = 125,504,258. [6] This is the difference between the number of locations eligible for the RAF in the first two support scenarios shown on the bar chart. These support scenarios use $50 and $55, respectively, for their target benchmarks, and different alternative technology cutoff levels to reach the equivalent fund size of $1.75 billion. The 1,142,544 locations covered by the RAF in the lower benchmark scenario include the 900,182 locations covered by the RAF in the higher-benchmark scenario, plus an additional 242,362 locations. These 242,362 locations are shifted into the CAF in the higher-benchmark scenario. [7] Given the significant impact that small shifts in the take rate can have on the number of locations covered by the CAF and the RAF and the amount of excess subsidy provided to CAF recipients, the Wireline Competition Bureau should base its take rate for estimating revenues on actual contemporary data about broadband adoption in previously unserved areas, rather than speculative estimates. Neither the Wireline Competition Bureau nor any other commenter has provided data that validates the blended average take rates included in the chart (50% to 80%) in this Virtual Workshop. While the ACA disagrees with the approach of using a blended average take rate to estimate revenues, if the FCC chooses to follow this approach, the Commission should at least base its recommended take rate on data on broadband adoption in previously unserved areas. The best benchmarks to use would be recent data on broadband adoption over a five-year period in previously unserved areas. [8] See Omnibus Broadband Initiative, The Broadband Availability Gap: OBI Technical Paper No. 1, at 45,

30 [9] The Pew survey question from which the data is drawn asks about home Internet adoption via wireline broadband (DSL, FTTH, cable), wireline narrowband (dial-up) and non-phone/tablet wireless (satellite, USB dongle, fixed wireless) Internet usage. It does not ask about mobile broadband via a phone or tablet. [10] See Trend Data (Adults), accessed June 3, [11] This estimate is based on the timelines set forth in the Report and Order detailing the challenge process to be used to finalize the list of areas eligible for CAF Phase II support and the process for carriers to accept state-level commitments. The challenge process includes a 45-day period for challenges to the status of a given area, and an additional 45-day period for rebuttals to these challenges. Once the Wireline Competition Bureau adjudicates on these challenges and rebuttals and finalizes the list of CAF-eligible areas, price cap carriers have 120 days to accept funding on a state-by-state basis. The total of these periods is 210 days ( ). Given the deadline for comments for the Virtual Workshop in question is June 18, 2013 (day 169 of the year, with 196 remaining), funding will not be awarded until after January 1, See Report and Order, DA (May 16, 2013), para 21, 24. [12] An alternative, theoretically ideal method to determine the support threshold would be to weight the ARPUs based on CAF-funded, rather than merely CAF-eligible, locations but this is nearly impossible in practice to calculate because the number of CAF-eligible locations is dependent on the support threshold. [13] The International Bureau of the FCC, in its third annual International Broadband Data Report, used voice services with unlimited local and long distance when benchmarking voice and double-play prices across 29 international markets. We follow that precedent for our benchmarking exercise, although a few price cap carriers do not appear to offer unlimited long distance minutes. See International Broadband Data Report, IB Docket No , GN Docket No , Third Report, 27 FCC Rcd 9884, 9904 (Int l Bur. 2012). These charts support the comments filed on behalf of the American Cable Association yesterday on Workshop Topic: Support Thresholds. 28

31 Robin Tuttle, counsel for ACS Question 1 Carriers accepting CAF Phase II support must build their network to provide voice and broadband services to all covered locations, not just the customers who subscribe to their service. Support calculations under the Model are based both on the estimated cost of delivering service to those customer locations and the amount of those costs that the carrier is expected to recover from customers who choose to subscribe to the offered service. Thus, some accommodation must be made for the fact that carriers will not receive revenue for all customer locations. If the Model estimates the cost of building a network, with costs calculated on one hundred percent of the covered locations, ACS agrees that, to reflect the carrier s prospects for cost recovery from customers, the support threshold should incorporate a suitable downward adjustment based on a realistic assessment of the portion of locations where consumers will not subscribe to the supported voice and broadband services, and therefore will not generate any revenue. Given the structure of the CACM s Support Module, adjusting down the lower benchmark is the cleanest method of incorporating a realistic take rate into the model. Question 2 The Bureau s efforts to predict a take rate over a period of time must realistically account for the build-out of the network to all covered locations over time to meet the five-year deadline with assumptions about growth in the take rate trailing those increasing build-out requirements. At the beginning of the CAF Phase II build-out period, broadband service meeting the Commission s CAF Phase II parameters is unlikely to be available throughout the CAF Phase II support area. Moreover, even when broadband is available and affordable, there will be a learning curve as consumers gradually become educated in the capabilities and benefits of broadband. Demographics of the served areas will also impact take rates. Another factor that varies according to the carrier receiving support, but that is still a critical factor in setting a take rate for the funding threshold is the starting point of a carrier s take rate prior to receiving CAF Phase II support. A lower starting point will require a more significant increase in subscribership over the supported period, as compared with other carriers. ACS submits that an 80% take rate is far in excess of a reasonably expected take rate for its service territory in Alaska. It greatly exceeds the current take rate of the largest broadband providers in the state. Even with the completion of a 4/1 Mbps network for supported locations in Alaska and with services offered in the proposed funding benchmark range of $40 to $50 per location per month, ACS believes the take rate for its voice and broadband services will remain substantially below the proposed 80% take rate. Based on the timing of network build-out, expected competitive pressures, and the realities of broadband adoption in Alaska, an average take rate of 80% is considerably overstated for ACS s service territory. 29